Journal of the Japanese Society for Sonic Arts, Vol.1 No.1 pp.17–21
Research Report
ALARM/WILL/SOUND: CAR ALARM SOUND PERCEPTION
EXPERIMENTS AND ACOUSTIC MODELING REPORT
Alexander Sigman
College of Music and Performing Arts
Department of Composition
Keimyung University
Daegu, South Korea
Nicolas Misdariis
Sound Perception and Design
Research and Development
STMS-IRCAM-CNRS-UPMC
Paris, France
ABSTRACT
car alarm be defined? To whom is the alarm directed: the
potential perpetrator, the car owner, or the public? StudThis article outlines the salient phases, goals, and reies summarized in the report cited above have indicated
sults of the sound perception and design research compothat even the most high-end audible alarms require a maxinent of alarm/will/sound, a multidisciplinary musical remum of ten minutes for professional car thieves to disable,
search project carried out in the context of the IRCAM
and in most instances fail to prevent break-ins. The owner
IRC (Interface Recherche-Création) Musical Residency
may or may not be within earshot of his/her respective car
Research program. After the rationale for and motivations
alarm when it is activated. However, given the homogenebehind the project are presented, the following research
ity of car alarm emissions, the alarm may not be detected
milestones are described: 1) a sound perception experiin time for the car owner to intervene. As a result of the
ment testing source typicality of a sub-category of sounds
sheer number of false alarms, as well as the aforemenwithin the corpus; 2) an acoustic descriptor space in which
tioned element of aural annoyance (among other factors),
a subset of the stimuli employed in the typicality experimembers of the more public have a greater documented
ment were situated; 3) the construction of synthetic autendency to flee from or simply ignore an activated alarm
ditory warnings from sound-sources within the descriptor
than to proceed towards the source.
space, prototypical environmental sound envelopes, and
In the presence of a car alarm, wherein lies the boundinter-onset intervals (IOI’s) derived from extant car alarms;
ary between the public space of the vehicle’s physical enand 4) the design and results of a second experiment pervironment and the private territory of car? An audible
taining to levels of repulsion vs. attraction to the synthetic
alarm designates a boundary beyond the car’s physical
auditory warnings. Finally, short-, mid-, and long-term
perimetera ”grey zone” that is often creatively explored,
objectives and directions for the project are discussed.
as is made evident by the countless videos of car alarm
dance routines posted to Youtube. 2
Our approach to audible car alarm system design has
1. INTRODUCTION
been guided and constrained by a fundamental question:
wherein lies the alarm’s identity? Does it consist in the
alarm/will/sound 1 is a tripartite collaboration between comhardware components and context (i.e., embedded in the
poser Alexander Sigman, IRCAM Sound Perception and
Design (SPD) team researcher Nicolas Misdariis, and Stuttgart- automobile), or in the sounds that it emits and the interaction protocol by which it operates? If the latter, is it
based product designer/visual artist Matthias Megyeri. This
possible to transform the alarm system from a mechanism
project was begun in January 2013 under the IRCAM IRC
of (ineffective) deterrence into one of engagement? That
(Interface Recherche-Création) Musical Research Resiis to say: through the expansion and customization of its
dency program, and is still in progress at present (Septemsonic vocabulary and potential modes of human-machine
ber 2014).
interaction, could this device be repurposed as a sort of
Taking as a point of departure the proven ineffectivevirtual instrument that the passerby (or car owner) learns
ness of current audible car alarm systems as deterrents [1]
to manipulate, with the help of audio-visual feedback?
and the relative lack of research into and development of
By the same token: could the alarm gain sensitivity to
audible car alarm design compared to other sound-emitting
more physical parameters than simply physical proximcomponents of vehicles (e.g., the audio system, engine,
ity? Could temporal variation in these physical paramehorn, turn signal, or door), we have sought to produce inters trigger time-varying sonic responses? At this point,
novative modified car alarm prototypes. The design of
it is worth mentioning that the aim of the study is notat
these prototypes would be informed by musical, artistic,
least, in the short termto address mnemonic and learning
scientific, and industry expertise, as well as sound percepissues attached to auditory warning stimuli. Even if these
tion research and acoustic modeling.
mechanisms could radically change the way people react
As an often-ignored and predictable source of noise
to alarms, our investigation do not pertain directly to how
pollution, the car alarm as an auditory warning device
raises a host of intriguing questions of a sociological na2 Here
ture. How may the essential functionality of the audible
is
a
particularly
well-choreographed
example:
1
http://www.youtube.com/watch?v=1Li3mNl2-EM
September 2014).
Also referred to as ”a/w/s” in this paper.
– 17–
(accessed
22
has consisted of constructing synthetic auditory
Indeed, it was not our intent purely to focus upon
warnings via the integration of the source-sounds within
enhancing the car alarm's deterrence effectiveness.
3
the descriptor space, prototypical auditory warning
Non-audible devices such as the Lojack system have
Journal of the Japanese Society for Sonic
Arts, Vol.1
No.1 pp.17–21 and the inter-onset intervals
temporal
morphologies,
been associated with a high documented vehicle
(IOI's) of real car alarm sounds. These synthetic
recovery rate. Rather than entirely replacing an existing
warnings
will then
beintegration
tested for
their
respective
people
integrateofover
time the
information induced
by an
auditory warnings
via the
of the
source-sounds
system
in service
security
enhancement,
the focus
capacities
for
repulsion
vs.
attraction
in
a second
alarm.
This
is
a
question
that
would
require
a
long-term
within
the
descriptor
space,
prototypical
auditory
warnhas been placed on expanding and enhancing this
experiment.
Concurrently,
Matthias
Megyeri
is
experimental
paradigm
to
accurately
and
rigorously
invesing
temporal
morphologies,
and
the
inter-onset
intervals
system's functionality and sonic potential. Despite the
tigate.
(IOI’s)
of
real
car
alarm
sounds.
These
synthetic
warnings
developing
hardware
designs
for
the
eventual
seemingly unique nature of the project and the
Indeed, it was not our intent purely to focus upon enare currently being
respectiveascapacities
for
prototypes,
whichtested
willforbetheir
exhibited
interactive
collaborative model underlying it, aspects of
hancing the car alarm’s deterrence effectiveness. Nonrepulsion
vs.
attraction
in
a
second
experiment.
Concurinstallations in public and gallery spaces during the
alarm/will/sound
aresuch
in as
facttheextensions
of Matthias
3
audible devices
Lojack system
rently,
Matthias
4have been
final
phase
of theMegyeri
project.is developing hardware designs
Megyeri's
workwith
ona high
domestic
securityvehicle
systems
andrate.
associated
documented
recovery
for the
eventual
prototypes,
will be exhibited
inGiven the breadth of which
alarm/will/sound,
we ashave
Alexander
compositional
interestssystem
in the
Rather Sigman's
than entirely
replacing an existing
in serteractive installations in public and gallery spaces during
decide
focus
in project.
the present article on the sound
influence
sonic phenomena
in physical
environments
vice of security
enhancement,
the focus
has been placed
the finaltophase
of the
corpus
elaboration
characterization
process
of
on
expanding
and
enhancing
this
system’s
functionality
on the aesthetics of the composer/sound artist and the
Given the breadth and
of a/w/s,
we have decide
to focus
Phase
I,
the
source
typicality
experiment
and
acoustic
and
sonic
potential.
in
the
present
article
on
the
sound
corpus
elaboration
and
impact that a composer/sound artist may have on
feature
modelingprocess
completed
during
and the
Despite the
the seemingly
nature of the
project
characterization
of Phase
I, thePhase
sourceII,typicality
transforming
physical unique
environment
(and
by and
5
synthetic
auditory
warning
construction
and
deterrence
the
collaborative
model
underlying
it,
aspects
of
a/w/s
are
experiment
and
acoustic
feature
modeling
completed
durextension, the human behaviors therein). The IRCAM
in fact extensions of Matthias Megyeri’s work on domestic
ing
Phase
II,
and
the
synthetic
auditory
warning
construcvs.
engagement
experiment
of
the
third
phase.
All
of
Sound Design and
Perception research team's
security systems 4 and Alexander Sigman’s compositional
tion and
deterrence
vs. engagement
experiment
of the
these
project
achievements
pertain
to
the
creation
of
involvement
theinfluence
automobile
industry
has assumed
interests with
in the
of sonic
phenomena
in physithird phase.and
All acoustic
of these project
achievements
pertain
to
semantic
classification
of
the
alarm
the form
of a partnership
with
Renaultofonthesound
design
cal environments
on the
aesthetics
composer/sound
the creationsounds.
of semantic
acoustic classification
of the
prototype
Theandclassification
methodologies
for electric
vehicles
(in collaboration
with composer
artist and
the impact
that a composer/sound
artist may
alarm prototype sounds. The classification methodologies
employed will be critical to both subsequent research
Andrea
a follow-up
on environment
electric vehicle
haveCera),
on transforming
thestudy
physical
(and by
employed will be critical to both subsequent research and
and
to exhibiting
the prototypes
in an interactive
art
5
extension,
human
behaviors therein).
detectabilty
in the
urban
environments
[2], and an earlier
to exhibiting
the prototypes
in an interactive
art installainstallation
context.
Sound
Design
Perception
tion context.
study onThe
car IRCAM
horn sound
quality
[3].and
Research
topicsresearch
in
team’s involvement
with theinteraction
automobilehas
industry
has asthe domain
of human-machine
included
sumed theofform
of afeatures
partnership
with Renault
on sound
the influence
audio
on perceived
urgency
design for electric vehicles (in collaboration with comand its application to car interior Human-Machine
poser Andrea Cera), a follow-up study on electric vehiInterfaces
[4] and inthe
influence
of naturalness
of
cle detectabilty
urban
environments
[2], and an earlier
auditory
feedback
of
an
interface
on
perceived
usability
study on car horn sound quality [3]. Research topics in the
and pleasantness
[5]. In addition,
Sigman's has
background
domain of human-machine
interaction
included the
in Cognitive
Science
and
timbre
perception
has been
influence of audio features on perceived urgency
and its
application
to car interior
Human-Machine
relevant
to the project's
collaborative
model. ItInterfaces
is thus [4]
the both
influence
naturalness
of auditory
feedback of
hopedandthat
the of
artistic
and research
outcomes
an interface onwill
perceived
usability
pleasantness
alarm/will/sound
contribute
notandonly
to the [5].
In addition,and
Sigman’s
background
Cognitive
Science
understanding
development
of invehicle
alarm
and
timbre
perception
has
been
relevant
to
the
project’s
systems specifically, but also to the design and
collaborative model. It is thus hoped that both the artistic
classification
of auditory
warnings
in general. will contribute
and research
outcomes
of alarm/will/sound
3
not only to the understanding and development of vehicle
http://www.lojack.com/Home
(accessed
March
2014).
alarm systems specifically,
but17also
to the
design and clas4
E.g., sification
Sweet Dreams
Security, awarnings
commercial
of security products
of auditory
inline
general.
developed
and
distributed
by
Megyeri:
http://www.sweetdreamssecurity.com/sweetdreamssecurity.html
(accessed 17 March
2. 2014).
PROJECT PHASES AND GOALS
5
Sigman's VURTRUVURT (2011) for prepared violin and live electronicsFor
and practical
down the bottle
(2012) the
for bass
flute,was
installation,
livefour
purposes,
project
dividedandinto
electronics—both members of the VURT cycle—reflect these interprimary
phases:
three
research
and
production
phases
(see
ests. Scores and recordings to both works may be found on the comFigure
1)
and
one
presentation
and
user
experience
docuposer's website: http://lxsigman.com/media/audio.htm (accessed 17
phase. The first phase was devoted to the proMarchmentation
2014).
Figure 1. The three primary project phases. "XP1"
Figure
1. Therefer
threetoprimary
project1"
phases.
”XP1” and
and "XP2"
"Experiment
and "Experiment
”XP2”
refer
to
”Experiment
1”
and
”Experiment
2,” re2," respectively.
spectively.
3. BACKGROUND
duction and characterization of a corpus of potential alarm
Besides
the distributed
IRCAM Sound
Perception
Design
team
Copyright:
© 2014
Alexander Sigman
and Nicolas
Misdariis.
This is an
article
under the
terms of theand
Creative
Commons
sounds.
Subsequently,
a sound
perception
experiment
in open-access
mentioned
above,
several
important
researches
Atttribution
Unported, which
permits
unrestricted
use, distribution, studies
and reproduction
in any
medium,
provided
the original
authorhave
and
soundLicense
source3.0identifiability
was
designed
and conducted
source are credited.
informed each stage of the project. Initially, a survey of
on a significant portion of the corpus, in order to focus
on acoustic properties, rather than exclusively to sound
causality. A subset of the stimuli used in this experiment was then placed within an acoustic features descriptor space. Phase III has consisted of constructing synthetic
historical audible alarm patents and current standards was
made. Sound corpus construction was guided both by
classic twentieth century approaches to timbral classification (e.g., Pierre Schaeffer’s Traité des objets musicaux
[6]), as well as more recent studies in environmental sound
categories (e.g., Houix et al, 2012 [7]). Formal components were also extracted from Olivier Claude’s 2006 thesis La recherche intelligente des sons [8], where taxonomies
of natural, animal, human, and object/machine sounds are
proposed, such that sounds within these taxonomies are
organized into limited sets of morphological, causal (physical), and semantic sub-categories.
In Ballas (1993) [9], acoustic, ecological, perceptual,
and cognitive factors that influence the identification of
3
http://www.lojack.com/Home (accessed 22 September 2014).
E.g., Sweet Dreams Security, a commercial line of
security products developed and distributed by Megyeri:
http://www.sweetdreamssecurity.com/sweetdreamssecurity.html
(accessed 22 September 2014).
5 Sigman’s VURTRUVURT (2011) for prepared violin and live electronics and down the bottle (2012) for bass flute, installation, and
live electronicsboth members of the VURT cyclereflect these interests.
Scores and recordings to both works may be found on the composer’s
website: http://lxsigman.com/media/audio.htm (accessed 22 September
2014).
4
– 18–
Journal of the Japanese Society for Sonic Arts, Vol.1 No.1 pp.17–21
current environmental sounds were evaluated. As is explained in Section 5.2, this study was particularly relevant
to the sound causality confidence rating protocol utilized
in Experiment 1.
The acoustic modeling step was largely informed by
several studies done on the definition of an exhaustive set
of acoustic featuresat first dedicated to musical sounds
(Peeters et al, 2002 [10]) and the identification of some
of these features that are best suitable for describing similarities and differences of environmental sounds.
The auditory warning construction was grounded on
the standard template defined by Patterson (1990) [11]
and, among others, the direction taken by Edworthy (2011)
[12] to extend the conception of the inner structure of
such signals was considered. Moreover, our own auditory
warning design process was also based on prototypes of
morphological profiles found out by Minard et al. (2010)
[13] from an environmental sound corpus.
Finally, existing auditory design in automobiles (Yamauchi et al, 2004 [14], Kuwano et al 2007 [15]), and
approaches to the synthesis of new auditory warnings in
military helicopters (Patterson 1999 [16]) and intensive
care units (Stanton and Edworthy, 1998 [17]) were also
taken into account in this process.
Figure 2. Sound corpus taxonomy, constructed in
January-February 2013.
5. EXPERIMENT 1: SOURCE IDENTIFIABILITY
OF INDUSTRIAL/MECHANICAL SOUNDS
4. SOUND CORPUS TAXONOMY AND SOURCES
The first phase of the project entailed the elaboration and
characterization of a sound corpus to apply to the modified car alarm prototypes. As is presented in Figure 2,
the sound corpus taxonomy consists of three primary categories: individual sounds, ”auditory scenes,” or sound
complexes, and real car alarm sounds (i.e., the standard
repertoire of six auditory warnings typical of audible car
alarm systems). The Individual Sound category is further divided into 1) Synthetic/Electroacoustic; 2) Vocal; 3)
Film Danger Icons; and 4) Industrial/Mechanical Sounds.
Further subdivisions were made along semantic/contextual
andparticular at the lowest levels of the taxonomyacoustic
lines.
Among the non-synthetic sounds in all three primary
categories, the majority were mined from existing sound
databases (e.g., SoundIdeas, Blue Box, Auditory Lab 6
and freesound.org). Under the Auditory Scenes rubric, a
series of field recordings of public spaces in Parisstreets,
the Forum Les Halles shopping concourse, the Centre Pompidou, metro stations, and train car interiorswere compiled
in February 2013 by Alexander Sigman and Matthias Megyeri. It is intended that the collection of field recordings be
expanded over time to include further site-specific entries.
Synthetic individual sounds were generated and edited
using such synthesis software as AudioSculpt, 7 Pure Data
(Pd), 8 SuperCollider, 9 and the Python-based concatenative synthesis program Audioguide. 10
6 http://www.psy.cmu.edu/ auditorylab/website/index/home.html (accessed 22 September 2014).
7 http://anasynth.ircam.fr/home/english/software/audiosculpt
8 http://puredata.info/
9 http://supercollider.sourceforge.net/
10 AudioGuide
was
developed
by
composer
Ben
Hackbarth,
and
is
obtainable
from
his
website:
http://www.benhackbarth.com/audioGuide/doc.html
(accessed
22
September 2014).
5.1. Experimental Objectives
The first sound perception experiment was designed in
the interest of determining levels of source identifiability of sounds within the corpus. Based upon the results
of the experiment, it would be possible to construct an
abstractness-iconicity scale across the corpus, as well as
to determine the salient semantic and acoustic attributes
of the sounds using empirical data. However, given the
size and scope of the catalogue, the selected stimuli were
limited to a subset of the Industrial/Mechanical category.
This category was chosen due to a) the number of subcategories and entries; and b) the range of source abstractness and ecological context relative to other Individual
Sound categories.
5.2. Methods and Materials
In order to obtain data from a broad range of subjects over
a relatively short period of time (ca. one month), the
experiment was conducted in an online, crowd-sourced
format. 11 Subjects were asked to listen to each stimulus, provide a brief description of sound causality, and
indicate a confidence rating of sound causality identification on a 1-5 Likert scale (see Figure 3). 12 The stimuli could be played back any number of times, but could
not be paused and resumed mid-file. Thirty-nine stimuli
were presented in MPEG-3 format. Every sub-category of
the Industrial/Mechanical category was represented by at
least one stimulus.
11 The experiment may be found at the following URL:
http://recherche.ircam.fr/equipes/pds/projects/asigman/causality/
src/EvaluationStartTrial.php (accessed 22 September 2014).
12 This experimental protocol was based on Ballas (1993), who found
a correlation between the measure of confidence of sound causality and
the more laborious causal uncertainty measure (Heu) [9]
– 19–
Journal of the Japanese Society for Sonic Arts, Vol.1 No.1 pp.17–21
stability for each participant. Subjects could not advance
stability
for each
participant.
Subjects
could
not session
advance
to All
the subjects
next
trial
until
the "Sound
Source
were
required
to complete
a Description"
trial
to
thewas
next
until
thethetrials
"Sound
Source
Description"
field
filled
out.
Once
experiment
completed,
consisting
of trial
three
practice
prior
to was
being
directed
field
filled
experiment
completed,
to
thewas
experiment,
inOnce
orderthe
totodetermine
stabilsubjects
were out.
requested
submit judgment
a was
questionnaire
subjects
were
to submit
anotquestionnaire
ity
for each
could and
advance to
pertaining
toparticipant.
therequested
subjects'Subjects
professional
educational
the
next trial
the ”Sound
Sourceused
Description”
field
pertaining
to until
theaudio
subjects'
professional
andduring
educational
backgrounds,
equipment
the
was
filled
out.
Once
the
experiment
was
completed,
subbackgrounds,
audio
equipment
used
during
the
experiment, and acoustics of physical environment
in
jects
to submit
a physical
questionnaire
experiment,
and
acoustics
whichwere
therequested
subjects
were oflocated
at environment
the pertaining
time ofin
to
the subjects’
educational
which
the subjects
wereandlocated
at backgrounds,
the feedback
time of
participating
inprofessional
the experiment.
In addition,
audio
equipment
used
during
the
experiment,
andfeedback
acousparticipating
the experiment.
In addition,
regarding thein experiment
was solicited.
The
whole
tics
of
physical
environment
in
which
the
subjects
were
listening test
lasted approximately
thirtyTheminutes.
regarding
the experiment
was solicited.
whole
located
at
the
time
of
participating
in
the
experiment.
In
Subjects were
paid forapproximately
their participation.
listening
test notlasted
thirty minutes.
addition, feedback regarding the experiment was solicited.
From anwere
objective
point
of
view,
the use of an online
Subjects
not paid
their
participation.
The
whole listening
testfor
lasted
approximately
thirty minprocedure
can
appear
to
be
either
a
blessing
curse:
as
From
an
objective
point
of
view,
use orofa an
online
utes. Subjects were not paid for theirthe
participation.
mentionedcan
above,
it allows
aeither
largeanumber
oforparticipants
procedure
appear
to
be
blessing
a
curse:
From an objective point of view, the use of an onlineas
representing
aappear
broad
range
professional
mentioned
above,
it allows
aoflarge
number of
participants
procedure can
to
be either
a blessing
orbackgrounds
a curse: as
and
geographic
locations
to
complete
the
experiment
in a
representing
a
broad
range
of
professional
backgrounds
mentioned above, it allows a large number of participants
limited
period
of
time.
On
the
other
hand,
it
may
induce
and
geographic
locations
to
complete
the
experiment
in a
representing a broad range of professional backgrounds
"noisy,"
unreliable
data
due
theexperiment
to
limited
period
oflocations
time.
Ontomainly
the
other
hand,
it inability
may induce
and
geographic
complete
the
in
directly
listener
lack
of
a"noisy,"
limitedmanage
period of
time.
On
theorother
it may in-to
unreliable
datafatigue
mainly
due hand,
theconcentration
inability
within”noisy,”
this
protocol.
results
presented
below to
in
duce
unreliable
data
mainly
due of
theconcentration
inability
directly
manage
listenerThe
fatigue
or lack
section
5.3
should
be
examined
in
light
of
these
directly
manage
listener
fatigue
or
lack
of
concentration
within this protocol. The results presented below in
limitations.
within
this
presented
belowofin these
secsection
5.3protocol.
should The
be results
examined
in light
tion
5.3
should
be
examined
in
light
of
these
limitations.
limitations.
Figure 3. Experiment 1 user interface.
5.3 Figure
Results3. Experiment 1 user interface.
Figure 3. Experiment 1 user interface.
Of the
ca. 100 visitors to the experiment website, twenty5.3
Results
four subjects began the experiment. Of this subject pool,
Of
thefifteen
ca. 100
visitorscompleted
to the experiment
twentyonly
subjects
all trials.website,
Data collected
four
thewho
experiment.
Of thisthesubject
5.3.
Results
fromsubjects
the ninebegan
subjects
did not reach
end ofpool,
the
only
fifteenwas
subjects
completed
trials. Data collected
experiment
excluded
from theallanalysis.
Of
the
ca.nine
100
towho
the
website,
twentyfrom
the
subjects
did
not reach
the end
the
Figure
4 visitors
indicates
the experiment
confidence
ratings
ofof the
four
subjects
began
the
experiment.
Of
this
subject
pool,
experiment
was
excluded
from
the
analysis.
fifteen subjects across the thirty-nine stimuli. The stimuli
only Figure
fifteen subjects completed
all trials. Data
collected
thefrom
confidence
ratings
of the
are indicated4 onindicates
the x-axis
left to right
in order
of
from
the
nine
subjects
who
did
not
reach
the
end
of
the
fifteen
subjects
across
the
thirty-nine
stimuli.
The
stimuli
confidence
rating
(from
low
to
high).
A
significant
experiment was excluded from the analysis.
are
indicated
on was
the x-axis
from
left lowest
to rightand
in order
of
difference
t-test
applied
to the
highest
Figure 4 indicates the confidence ratings of the fifteen
confidence
ratingratings
(frominlow
totohigh).
A
significant
mean
confidence
order
locate
the
threshold
subjects across the thirty-nine stimuli. The stimuli are indifference
t-test
was
applied
to
the lowest
and
betweenoniconic
(strongly
identifiable)
non-iconic
dicated
the
x-axis
from
left
to right
in and
order
of highest
confimean
confidence
ratings
in
order
to
locate
the
threshold
sounds.
Stimuli
and
their
respective
mean
confidence
dence rating (from low to high). A significant difference
ratingswas
areiconic
listed (alphabetically)
in
Figure
5. mean
between
(strongly
identifiable)
and
non-iconic
t-test
applied
to the lowest
and
highest
consounds. ratings
Stimuliinand
respective
mean confidence
fidence
ordertheir
to locate
the threshold
between
ratings(strongly
are listedidentifiable)
(alphabetically)
in Figure 5.
iconic
and non-iconic
sounds. Stimuli and their respective mean confidence ratings are listed
(alphabetically) in Figure 5.
Figure 4. Confidence rating means () and standard
deviations
for 39rating
stimuli
indicated
fromand
lowstandard
(left)
Figure4.4.()
Confidence
rating
means
Figure
Confidence
means
and ()
standard
deviatodeviations
high
tions
for(right).
thirty-nine
indicated
from from
low (left)
to
() forstimuli
39 stimuli
indicated
low (left)
high
(right).
to high
(right).
Stimulus
Mean Confidence Rating
Amongst
provided by
the
1Stimulus
airco_offthe sound source descriptions
2.80
Mean Confidence
Rating
subjects
(in
the
field
below
the
confidence
scale
on
the
21 airco_on
3.73
airco_off
2.80 ranged in confidence
experiment
interface), the responses
32 airplane_beginning
4.60
airco_on (In one case, for
3.73
and
specificity.
instance, a subject iden4
airplane_end
3.00
3
airplane_beginning
tified the brand of metronome of4.60
one of the stimuli.)
54 alarm_clock_bell
4.67
airplane_end
3.00
65 ball_ricochet
3.60
alarm_clock_bell
4.67
5.4. Discussion
76 boat_motor_edited
2.80
ball_ricochet
3.60
87 can_crushed_edited
3.00
The
relatively
high attrition rate
amongst potential subboat_motor_edited
2.80
jects
(100
to
twenty-four
to
fifteen)
98 can_knocked_over
3.20
can_crushed_edited
3.00 suggests two factors
that
discouraged
participants from
continuing with the ex10
cannon_shot
4.13
9 can_knocked_over
3.20
periment:
1) the duration required
to complete the task
11
circular_saw
3.67
10 cannon_shot
4.13 nature of the sounds
(45-60
minutes)
and
2)
the
fatiguing
12
4.87
11 corkscrew
circular_saw
3.67
presented.
There was also no available
”Save and Con13
electric_screwdriver
3.33
12 corkscrew
tinue”
option, so the experiment4.87
had to be conducted in
14
fog_horn
4.33
one
Experiment 2 requires
13 sitting.
electric_screwdriver
3.33 a shorter completion
15
food_processor_off
2.00 of stimuli of shorter
time
15-20 minutes), and consists
14 (ca.
fog_horn
4.33
16 grandfather_clock_bells 4.87
durations.
15 food_processor_off
2.00
17 The
grenade_blast
individual differences 2.93
in4.87
confidence ratings and
16 grandfather_clock_bells
specificity
of responses may be3.20
attributed in part to differ18 hand_mixer_off
17 grenade_blast
2.93
ences
in experience with mechanical/electrical
tools and
19 helicopter_hovering
4.47
18
hand_mixer_off
3.20
industrial
equipment. The sounds
20 helicopter_passing
5.00 that caused the most
19 helicopter_hovering
4.47
confusion
either had contextually-obscure
sources (e.g.,
21 machine_gun_3_iteration
4.33
20
helicopter_passing
5.00
wooden
gears
or
milling
machine),
impulsive
(e.g., grenade
22 marbles_in_vase
2.67
blast,
marbles in vase, or a ball-ricochet)
or were steady21 machine_gun_3_iteration
4.33
23 med_clock_ticks
2.83
state
drone-like in nature, with
motor sources (e.g., air
22 and
marbles_in_vase
2.67
24
metronome
3.60 motor, or food procesconditioner,
microwave oven, boat
23
med_clock_ticks
2.83
25 microwave_oven_begin
3.07 context-specific (e.g.,
sor).
The
iconic sounds were quite
24 microwave_oven_end
metronome
3.60
26
3.80
grandfather
clock bells, antique
train, or machine gun),
25
microwave_oven_begin
3.07
27 milling_machine_on
time-varying
(e.g. helicopter, 2.13
airplane, or train passing),
26common
microwave_oven_end
3.80 a bottle or alarm
or
(e.g., corkscrew 3.93
opening
28
rachet
27
milling_machine_on
2.13 that were segmented
clock).
Interestingly, a few sounds
29 sander_off
3.27
28
rachet
3.93 confidence scale ratinto
two stimuli produced different
30 sander_on
3.73
ings
for
each
segment.
The
sound
29 saw_cutting_pipe
sander_off
3.27 ”airco-off” (air con31
4.47
ditioner
off)
had
a
mean
confidence
30 shaver_middle
sander_on
3.73 rating of 2.8, while
32
4.13
”airco-on”
was
correlated
with
a
mean rating of 3.73. Sim31 stopwatch_beep
saw_cutting_pipe
4.47
33
4.67
ilarly,
”airplane-end”
had
a
mean
rating of 3.0, while the
32 train_antique
shaver_middle
4.13
34
rating
”airplane-beginning” fell4.53
to the right of the iconic33 train_rail_noise
stopwatch_beep
4.67
35
4.60
ity
threshold, at 4.6. As otherwise
drone-like, steady-state,
34
train_antique
4.53
36
vaccuum_end
3.60
motor-produced sounds, the onsets
of ”airco-on” and ”airplane35 vacuum_begin
train_rail_noise
4.60 may have provided
37
3.93
beginning,”
as well as ”sander-on,”
36 vacuum_cleaner_in_motion
vaccuum_end
3.60 offsets. By contrast,
more
spectral
cues than the respective
38
2.33
the
of ”microwave-oven-end”
(confidence rat37termination
vacuum_begin
3.93
39
wooden_gears_excerpt
2.00
ing
to have given
more
than
”microwave38= 3.8)
vacuum_cleaner_in_motion
2.33cues
Figure
5.seems
The thirty-nine
stimuli
(listed
alphabetically)
oven-begin”
(confidence
rating
=
3.07).
39 their
wooden_gears_excerpt
2.00 ratings.
and
respective mean confidence
Despite5.the
lower
levels ofstimuli
participation
expected,
Figure
The
thirty-nine
(listedthan
alphabetically)
these
results
did
enable
us
to
construct
an
abstractionand their respective mean confidence ratings.
iconicity scale and to determine salient semantic charac-
Copyright: © 2014 Alexander Sigman and Nicolas Misdariis. This is an open-access article distributed under the terms of the Creative Commons
Atttribution License 3.0 Unported, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and
source are credited.
20–
Copyright: © 2014 Alexander Sigman and Nicolas Misdariis. This is–an
open-access article distributed under the terms of the Creative Commons
Atttribution License 3.0 Unported, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and
source are credited.
an online
a curse: as
participants
ackgrounds
riment in a
may induce
nability to
ncentration
below in
of these
ite, twentyubject pool,
a collected
end of the
ngs of the
The stimuli
in order of
significant
nd highest
e threshold
non-iconic
confidence
entries.
Figure 4.(100
Confidence
rating means
() and
standard
subjects
to twenty-four
to fifteen)
suggests
two
Several approaches were considered, including Mel
deviations
()
for
39
stimuli
indicated
from
low
(left)
factors that discouraged participants from continuing with
Frequency
Cepstral Coefficient (MFCC) attached to a
to high
(right). 1) the duration required to complete the
the
experiment:
Journal of the Japanese Society for Sonic
Arts, Vol.1 Mixture
No.1 pp.17–21
Gaussian
Model (GMM) and a Multitask (30-45 minutes) and 2) the fatiguing nature of the
Dimensional Scaling (MDS) analysis. It was ultimately
sounds presented. There was also no available "Save and
decidedCepstral
that weighted-mean
Perceptual
Spectral
Stimulus option, so the experiment
Mean Confidence
Rating
quency
Coefficient (MFCC)
attached
to Centroid
a GausContinue"
had to be conducted
(PSC)
and Model
Harmonicity-to-Noise
Ratio (HNR)—two
1
airco_off
2.80
sian
Mixture
(GMM)
and
a
Multi-Dimensional
Scalin one sitting. Experiment 2 will require a shorter
acoustic
parameters
included
in the decided
IrcamDescriptor
2.7
ing
(MDS)
analysis.
It
was
ultimately
that
weighted2
airco_on
3.73
completion time (ca. fifteen minutes), and will consist of
toolbox
[19]—would
be
employed
based
upon
a
2010
mean Perceptual Spectral Centroid (PSC) and Harmonicity3 airplane_beginning
stimuli
of shorter durations. 4.60
study byRatio
Misdariis
et acoustic
al. on parameters
metadescription
to-Noise
(HNR)two
includedand
in
4 The
airplane_end
3.00
individual differences in confidence ratings and
modeling
of
environmental
sounds
such
as
interior
car
the
IrcamDescriptor
2.7
toolbox
[18]would
be
employed
5 alarm_clock_bell
specificity
of responses may4.67
be attributed in part to
based
upon
2010 study by
et al.
metadesounds,
air aconditioners,
carMisdariis
horns, and
car on
doors
[11].
6 ball_ricochet
differences
in experience with3.60
mechanical/electrical tools
scription
andofmodeling
of environmental
suchthat,
as
This choice
descriptors
was made due sounds
to the fact
7
boat_motor_edited
2.80
and industrial equipment. The sounds that caused the
interior
car
sounds,
air
conditioners,
car
horns,
and
car
in
the
Misdariis
et
al.
study,
it
has
been
shown
that
these
8 can_crushed_edited
3.00
most
confusion either had contextually-obscure
sources
doors
[19]. This
choicetoofbe
descriptors
wasinmade
due to the
two features
appear
significant
the perceptual
9
can_knocked_over
(e.g., wooden gears or milling3.20
machine), impulsive (e.g.,
fact
that, in the
Misdariis
et al. study,
it has
been shown
description
of
environmental
sounds.
Moreover,
this
10 cannon_shot
grenade
blast, marbles in vase,4.13
or a ball-ricochet) or were
that
these
twoindicated
features appear
to abefirst-level
significantdescription,
in the perstudy
also
that,
in
11 circular_saw
steady-state
and drone-like in3.67
nature, with motor sources
ceptual
description
of environmental
Moreover,if
considering
the mean
value of thesounds.
features!even
12 corkscrew
4.87 oven, boat motor, or
(e.g.,
air conditioner, microwave
this study also indicated that, in a first-level description,
sounds are objectively
non-stationary!corresponds
to a
13 electric_screwdriver
food
processor). The iconic 3.33
sounds were quite contextconsidering
the mean value
of the featureseven if sounds
perceptually
relevant
process
(see
correlation
values
14
fog_horn
4.33
are objectively non-stationarycorresponds to a perceptuspecific (e.g., grandfather clock bells, antique train, or
between
acoustic
and perceptual
dimensions
in
15 food_processor_off
2.00 helicopter, airplane, or
ally
relevant
processfeatures
(see correlation
values between
acousmachine
gun), time-varying (e.g.
[11]).
That and
is the
primary dimensions
reason that inwe[19]).
choseThat
in the
16
grandfather_clock_bells
4.87
tic
features
perceptual
is
train passing), or common (e.g., corkscrew opening a
present
study
to remain
on chose
this first
level
and construct
17 grenade_blast
2.93
the
primary
reason
that we
in the
present
study to
bottle
or alarm clock). Interestingly,
a few sounds that
our
“Acoustic
Features
Space”
upon
mean
values ofFeathe
18 hand_mixer_off
remain on this first level and construct our Acoustic
were
segmented into two 3.20
stimuli produced different
features.
tures
Space
upon
mean
values
of
the
features.
19
helicopter_hovering
4.47
confidence scale ratings for each segment. The sound
two-dimensional
PSC-HNR
space
calculated
AA
two-dimensional
PSC-HNR
space
waswas
calculated
and
20 helicopter_passing
5.00had a mean confidence
"airco-off"
(air conditioner off)
and
populated
by
the
thirty-one
Industrial/Mechanical
populated
by
the
thirty-one
Industrial/Mechanical
sounds
21
machine_gun_3_iteration
4.33
rating of 2.8, while "airco-on" was correlated with a mean
soundsin Experiment
tested in Experiment
1 that
fell below
the
tested
1 that fell below
the iconicity
thresh22 marbles_in_vase
2.67
rating
of 3.73. Similarly, "airplane-end"
had a mean
old.
Moreover,
in order
to fill empty
or underrepresented
iconicity
threshold.
Moreover,
in
order
to
fill
empty
or
23
med_clock_ticks
2.83
rating of 3.0, while the rating "airplane-beginning" fell to
zones
of
the
space,
hybrid
sounds
were
constructed
via
the
underrepresented
zones
of
the
space,
hybrid
sounds
were
24 right
metronome
3.60 at 4.6. As otherwise
the
of the iconicity threshold,
constant
cross-synthesis
of pairs
of one-second of
samples
constructed
via the constant
cross-synthesis
pairs of
of
25 microwave_oven_begin
3.07
drone-like,
steady-state, motor-produced
sounds, the
extant
sounds
in AudioSculpt.
The resulting
acoustic feaone-second
samples
of
extant
sounds
in
AudioSculpt.
The
26
microwave_oven_end
3.80
onsets of "airco-on" and "airplane-beginning," as well as
tures
space
is illustrated
in space
Figureis6.illustrated in Figure 6.
resulting
acoustic
features
27 milling_machine_on
2.13more spectral cues than
"sander-on,"
may have provided
28 respective
rachet
3.93
the
offsets. By contrast,
the termination of
29
sander_off
3.27 rating = 3.8) seems to
"microwave-oven-end" (confidence
30 sander_on
3.73"microwave-oven-begin"
have
given more cues than
31 saw_cutting_pipe
(confidence
rating = 3.07). 4.47
32 Despite
shaver_middle
4.13 of participation than
the lower levels
33
stopwatch_beep
expected, these results did 4.67
enable us to construct an
34 train_antique
4.53 to determine salient
abstraction-iconicity
scale and
35 train_rail_noise
semantic
characteristics of 4.60
the sounds tested. It was
36 vaccuum_end
3.60
concluded
that the sounds falling
to the right of the
37 vacuum_begin
iconicity
threshold would not3.93
function effectively in the
38 alarm
vacuum_cleaner_in_motion
2.33 too closely associated
car
context, as they were
39 wooden_gears_excerpt
2.00 very well exist within a
with
specific sources, which may
Figure
5.
The
thirty-nine
stimuli
(listed alphabetically)
vehicle's
immediate
environment
(e.g.,
and
their
respective
mean
confidence
ratings.
Figure
5.
The
thirty-nine
stimuli
(listed
alphabetically)
trains/airplanes/helicopters and clock
chimes).
and their respective mean confidence ratings.
ariis. This is an open-access article distributed under the terms of the Creative Commons
icted use, distribution,
and reproduction
in any medium,
original author
teristics
of the sounds
tested. provided
It was the
concluded
that and
the
Figure 6. Perceptual Spectral Centroid (PSC)Figure
6.
Perceptual Ratio
Spectral
(PSC)sounds falling to the right of the iconicity threshold would
Harmonicity-to-Noise
(HNR)Centroid
acoustic features
Harmonicity-to-Noise Ratio (HNR) acoustic features
not function effectively in the car alarm context, as they
Copyright:
© 2014 Alexander
Sigman
Nicolas sources,
Misdariis. which
This is an open-access
article distributed
under the terms of the Creative
space containing
industrial/mechanical
source Commons
sounds
were
too closely
associated
withand
specific
Atttribution
License
3.0 within
Unported,
which permits
unrestricted
use, distribution,
and reproduction
in any
provided (red
the original
and
(black
dots), actual
car medium,
alarm sounds
dots),author
and hymay
very well
exist
a vehicle’s
immediate
environsource
are
credited.
brid sounds (blue dots).
ment (e.g., trains/airplanes/helicopters and clock chimes).
6. ACOUSTIC MODELING: PSC-HNR
DESCRIPTOR SPACE
The next step in the corpus characterization process was
to compute perceptually relevant acoustic descriptors. The
objective was to construct an acoustic descriptor space in
which to situate the corpus sounds, thereby enabling us
to trace relative distances between constituent sounds, as
well as between extant sounds and new entries.
Several approaches were considered, including Mel Fre– 21–
7. SYNTHETIC AUDITORY WARNING
CONSTRUCTION
Synthetic auditory warnings were constructed by combining: a) selected industrial/mechanical source sounds and
hybrids placed within the acoustic descriptor space described above; b) five typical environmental sound envelopes; and c) the inter-onset intervals (IOI’s) of the six
alarms comprising a standard car alarm repertoire.
536,
ms., respectively).
respectively).
In addition,
sounds and hybrids placed within the acoustic descriptor
536, and
and 2082
2082 ms.,
In addition,
it is it is
space described
above; above;
b) fiveb) typical
environmental
space described
five typical
environmental
possible
trigger a a Morphology
Morphology
Sequencer,
possible to
to trigger
Sequencer,
whichwhich
sound envelopes;
the inter-onset
intervals
(IOI's)
sound envelopes;
and c) and
the c)
inter-onset
intervals
(IOI's)
cycles through
through the
at at
thethe
current
IOI rate.
cycles
thesix
sixBPF's
BPF's
current
IOI The
rate. The
Journal
of the
Society for Sonic
Arts, Vol.1Sequencer
No.1 pp.17–21
of the
six alarms
comprising
a standard
car
alarm
of the six
alarms
comprising
a standard
carJapanese
alarm
Morphology
loops
until
it isitdeactivated.
Morphology
Sequencer
loops
until
is deactivated.
repertoire.repertoire.
Of the set of stimuli shown in Figure 6, six original
Of the set
stimuli
shownshown
in Figure
6, six6, original
Of of
the
of cross-synthesized
stimuli
in hybrids
Figure
sounds
andset
three
thatsix
lieoriginal
at the
sounds
and
three
cross-synthesized
hybrids
thatat
liethe
at the
sounds and
three
cross-synthesized
hybrids
thatwere
lie
extremes and center of the descriptor space
selected.
extremes
andofcenter
of the descriptor
space
were
selected.
extremes and
center
the
descriptor
space
were
selected.
In the Minard, et al. study [14] mentioned previously, six
In the et
Minard,
et al. study
[14] mentioned
previously,
In the Minard,
al. study
mentioned
previously,
six six
perceptually
distinct[14]
environmental
sound
morphologies
perceptually distinct environmental sound morphologies
weredistinct
devised and tested: 1) sound
stable; 2) decreasing; 3)
perceptually
were devised environmental
and tested: 1) stable; 2)morphologies
decreasing; 3) inincreasing;
4)
pulse-train;
5)
single
impulse;
were devised
and
tested:
1)
stable;
2)impulse;
decreasing;
3) 6)
creasing; 4) pulse-train; 5) single
and 6)and
rolling
rolling
(see
Figure
7).5)
Given
the iterative
these
increasing;
pulse-train;
single
impulse;
andofalarms,
6)
(see4)Figure
7). Given
the
iterative
nature nature
of these
alarms,
the category
category
wasasexcluded,
as recogthis
theFigure
”stable”
this inhibits
rolling (see
7)."stable"
Givenwas
theexcluded,
iterative
nature
of these
inhibits
recognition
of
new
iterations
in
an
ecological
or
nition
of new iterations
an ecological
or experimental
alarms, the
"stable"
categoryinwas
excluded,
as this
experimental
context.
context.
inhibits recognition of new iterations in an ecological or
experimental context.
Figure 8. Six car alarm inter-onset intervals (IOI's).
For each profile, the blue line represents the whole
profile 8.
andSix
the car
red alarm
line represents
theintervals
elementary
Figure
inter-onset
(IOI's).
profile
used
for
computing
standard
values
of
Figure 8. Six car alarm inter-onset intervals IOI's.
(IOI’s). For
For each profile, the blue line represents the whole
each profile,
the the
wholeelementary
profile
profile
and the
the blue
red line
linerepresents
represents
and
the red line represents
the
elementary profileVS.
used for
8.
EXPERIMENT
2:
ATTRACTION
profile
used for computing standard values of IOI's.
computing standard values of IOI’s.
Figure 7. Six environmental sound envelopes, rendered
Figure
7. Sixfunctions
environmental
envelopes, rendered
as
breakpoint
(BPF's) sound
in Max/MSP.
REPULSION TO SYNTHETIC AUDITORY
as breakpoint functions (BPF’s) in Max/MSP.
WARNINGS
As is
presented in Figure
8, the
inter-onsetrendered
intervals of
Figure 7. Six
environmental
sound
envelopes,
8. EXPERIMENT 2: ATTRACTION VS.
14
six
standard
car
alarms
were
calculated
(in
Matlab
).
8.
EXPERIMENT
2: ATTRACTION VS.
as breakpoint functions (BPF's) in Max/MSP.
AUDITORY
As wasREPULSION
the case forTO
Experiment
1, Experiment
2 will be
Given
thepresented
similarityinin
IOI of
1 and 2,intervals
3 and 5,of REPULSION
As is
Figure
8, alarms
the inter-onset
TOSYNTHETIC
SYNTHETIC
AUDITORY
13
conducted
via
an
online,
crowd-sourced
format
on the
and
4 and 6 (respectively)
to each
other, (in
each
pair of IOI's
six standard
car alarms were
calculated
Matlab
).Given
WARNINGS
WARNINGS
As is presented
in Figure
8,alarms
the
inter-onset
of
IRCAM
Sound
Perception
and
Design
research
team
site.
the similarity
in
IOI of
1 and 2,IOI
3 intervals
and
5, and
4
and
was
averaged,
producing
three
distinct
durations:
300
14
six standard
car alarms
calculated
(inms.
). av- Subjects
will case
be for
presented
with 1,synthetic
auditory
6 (respectively)
towere
each
other,
each2082
pair
ofMatlab
IOI’s
was
ms.
(short),
536 ms.
(medium),
and
(long).
As was the
Experiment
Experiment
2 is bewasconducted
the
case at
forthe
Experiment
1,via
Experiment
2 will be
Given theeraged,
similarity
IOIthree
of alarms
1IOIand
2, 3 times
and300
5,five
warning
stimuli,
and
asked
to rate
the
level
attraction
distinct
durations:
ms. As
Since producing
the in
combination
of nine
sources
ing
present
time
an of
online,
crowdconducted
via
an
online,
crowd-sourced
format
or
repulsion
to
the
sound
on
a
three-point
scale
(short),
536
ms.
(medium),
and
2082
ms.
(long).
times three
IOI's other,
would still
many
and 4 andenvelopes
6 (respectively)
to each
eachproduce
pair oftoo
IOI's
sourced format on the IRCAM Sound Perception and on
De- the
14
(attraction/indifference/repulsion).
In
order
to
facilitate
sign
research
team
site.
Subjects
are
presented
with
stimuli
to
be
realistically
tested
in
a
sound
perception
was averaged,
producing
three distinct
durations:
Since
the combination
of nineIOI
sources
times 300
five en- IRCAM Sound Perception and Design research team site.
this
task, the
instructions
to the
experiment
will to
include
synthetic
auditory
stimuli,
asked
rate
the
experiment,
and
combinations
wouldtoocreate
will
be warning
presented
withandsynthetic
auditory
ms. (short),
536 ms.
(medium),
and
2082still
ms.produce
(long).
velopes
times
threeseveral
IOI’s
would
many Subjects
metaphors
to
other
sensory
modalities
(e.g.,
attraction
vs.
level
of
attraction
or
repulsion
to
the
sound
on
a
threecontradictions
among
the
parameters
(and
by
extension,
to be realistically
testedsources
in a soundtimes
perception
stimuli, and asked to rate the level of attraction
Since stimuli
the combination
of nine
five ex- warning
to (attraction/indifference/repulsion).
odors). In addition, the subjects In
will
be to
point scale
order
lose
perceptual
saliencecombinations
in an auditory
warning
context),
periment,
and IOI's
several
would
create
contra- orrepulsion
repulsion
to
the
sound
on toathe
three-point
scale
envelopesfurther
times three
would
still
produce
too
many
facilitate
this
task,
the
instructions
experiment
presented
with
the
following
ecological
scenario:
one
is inmade (and
on an
intuitive (but
dictions exclusions
among the were
parameters
by extension,
lose (attraction/indifference/repulsion). In order to facilitate
stimuli toempirical)
be realistically
in new
a sound
perception
clude metaphors
to other
sensory
modalities
attracconfronted
with a box
emitting
a given
synthetic(e.g.,
auditory
basis.
Fortested
onsets
of
sounds with
perceptual
salience
ininstance,
an auditory
warning
context),
furthis
task,
the
instructions
to
the
experiment
will
include
tion
vs.
repulsion
to
odors).
warning.
Does
one
feel
compelled
to
pick
up
and
open
experiment,
and
several
combinations
would
create
ather
decreasing
envelope
and
a
short
IOI
were
deemed
exclusions were made on an intuitive (but empirical)
In
the
interest
of
limiting
the
required
completion
time vs.
the box, ignore
it, orsensory
to run away?
With this
scenario
in
metaphors
to other
modalities
(e.g.,
attraction
imperceptible.
On the
opposite
of by
the extension,
scale,a decreassinglecontradictions
the
parameters
(and
basis. among
For instance,
new
onsets end
of
sounds
with
for
the
experiment
to
15-20
minutes,
forty-five
synthetic
mind,
subjects
will
complete
a
brief
listening
test/practice
impulse
envelopes
with
long
would
lose the
repulsion to odors). In addition, the subjects will be
ing envelope
andin
a short
IOI
wereIOI's
deemed
imperceptible.
lose perceptual
salience
an
auditory
warning
context),
auditory
warnings,
rathertothan
all possiblescenario:
combinations
trial
prior to
being
the ecological
experiment.
qualities
of were
an end
auditory
warning,
the envelopes
latency
On the opposite
of the scale,
presented
with
thedirected
following
one is
further exclusions
made
on single-impulse
an given
intuitive
(but
of the
sources,
temporal
morphologies,
and IOI’s
described
In the
interest
of limiting
the required
completion
between
onsets.
Similarly,
impulsive
and granular
sourcewith
long
IOI’s
would
lose
the
qualities
of
an
auditory
confronted
with
a
box
emitting
a
given
synthetic
auditory
empirical) basis. For instance, new onsets of sounds with
above
These stimuli
will
time
for are
the employed
experimentastostimuli.
fifteen minutes,
a subset
of be
thesesounds
wooden
gears,between
or a toppled
can) would
warning,(e.g.,
given
the latency
onsets.tinSimilarly,
imwarning.
Does
one
feel
compelled
toabove
pick
up
and
a decreasing
envelope
and
a
short
IOI
were
deemed
lected
on
the
basis
of
contrast
of
spectral
and
temporal
synthetic
auditory
warnings
described
will
be open
pulsive
and granular
source-sounds
(e.g., wooden gears,
not
effectively
be paired
with long IOI's.
the
box,
ignore
it,
or
to
run
away?
With
this
scenario
in
characteristics.
It isThese
hoped
that the
of this
imperceptible.
Onthese
thetinrestrictions
opposite
end
the scale,
singleemployed
as stimuli.
stimuli
willresults
be selected
on experthe
or Once
a toppled
can) would
notof
effectively
be possible
paired with
were
applied,
it was
to
iment
will
enable
us
to
determine
which
synthetic
audicomplete
brief listening
test/practice
basis subjects
of contrastwill
of spectral
anda temporal
characteristics.
impulse generate
envelopes
withof contrasting
long IOI'sstimuli
would
the in mind,
long
IOI’s.
an array
to be lose
employed
warnings
would
be more
effective
as deterrents,
It tory
isprior
hoped
that
the
results
of to
this
experiment
will enableand
trial
to
being
directed
the
experiment.
2. This
was
achieved
via
the
Max
6
patch.
In
qualities Experiment
of Once
an these
auditory
warning,
given
the
latency
restrictions were applied, it was possible to
which
could be
utilized
as auditory
feedback
in would
the conto
which
auditory
warnings
Indetermine
the interest
ofsynthetic
limiting
the required
completion
between onsets.
granular
source- in ustext
generateSimilarly,
an array ofimpulsive
contrastingand
stimuli
to be employed
of user interaction with the alarm prototypes.
14
http://www.mathworks.com/products/matlab/
Experiment
This was
viatin
the can)
Max 6would
patch. In time for the experiment to fifteen minutes, a subset of the
sounds (e.g.,
wooden2. gears,
or achieved
a toppled
the
patch,
the
six
environmental
sound
envelopes
are
Copyright:
©
2014
Alexander
Sigman
and
Nicolas
Misdariis.
This renis an open-access
distributedwarnings
under the terms
of the Creative
Commons
syntheticarticle
auditory
described
above
will be
not effectively be paired with long IOI's.
Atttribution
License 3.0 Unported,
which
permits unrestricted
use, distribution,
and reproduction in any
medium,
provided
the original author and
dered
as
breakpoint
functions
(BPF’s),
as
is
indicated
in
9.
FUTURE
WORK
employed as stimuli. These stimuli will be selected on the
Once these
restrictions were applied, it was possible to
source
Figureare7.credited.
The user first chooses from amongst the nine basis of contrast of spectral and temporal characteristics.
generate an
array
of
contrasting
stimuli
to
be
employed
in
possible source sounds mentioned above. This sound is
Goals for the short-term include completing Experiment
that thetheresults
this experiment
Experiment
2. This with
was one
achieved
viaBPF’s,
the Max
6 patch.
modulated
of the six
selected
from aIn
drop- It is2 hoped
and analyzing
results.ofThereafter,
we will will
focusenable
on
us
to
determine
which
synthetic
auditory
warnings
would
down menu. The BPF duration and triggering/looping rate
the interactivity component of the project, and determine
14
http://www.mathworks.com/products/matlab/
may be altered by clicking on one of nine IOI durations:
the optimal hardware and software development environsix corresponding to the IOI’s of the standard car alarms,
ments in which to pursue this. Once our alarm protoCopyright: ©
Alexander
Sigman and Nicolas
This is an
distributed
under thestage
termsof
ofdevelopment,
the Creative Commons
the2014
three
others corresponding
to theMisdariis.
aforementioned
av-open-access
typesarticle
have reached
a sufficient
they
Atttribution eraged
License inter-onset
3.0 Unported,
which permits
unrestricted
distribution,
and reproduction in any medium, provided the original author and
intervals.
In addition,
it is use,
possible
to
will be exhibited primarily in site-specific public-space
source are credited.
trigger a Morphology Sequencer, which cycles through
contexts, but also in gallery spaces. The establishment
the six BPF’s at the current IOI rate. The Morphology
of collaborative relationships with industry partners is in
Sequencer loops until it is deactivated.
progress, and should be confirmed within the next year or
so.
14
13
http://recherche.ircam.fr/equipes/pds/projects/asigman/deterrence/
src/EvaluationStart.php (accessed 22 September 2014).
http://www.mathworks.com/products/matlab/
– 22–
Journal of the Japanese Society for Sonic Arts, Vol.1 No.1 pp.17–21
[7] O. Houix, G. Lemaitre, N. Misdariis, P. Susini, and
I. Urdapilleta, ”A Lexical Analysis of Environmental Sound Categories,” Journal of Experimental Psychology: Applied, vol. 18, no. 1, pp. 52-80, 2012.
[8] O. Claude, La recherche intelligente des sons.
(Masters Thesis.) Universit de Provence, Aubagne,
France, 2006.
[9] J.A. Ballas, ”Common Factors in the Identification of
an Assortment of Everyday Sounds,” Journal of Experimental Psychology: Human Perception and Performance, vol. 19, no. 2, pp. 250-267, 1993.
Figure 9. Experiment 2 user interface.
When the prototypes are presented to the public, the
interaction models described previously will be featured.
Exhibition visitors will be able to trigger alarm sounds remotely via control stations or mobile devices, in a searchable catalog organized and characterized based upon the
results obtained from the two experiments and the constructed acoustic descriptor space. User experiences with
the various modes of interaction will be documented via
video/camera images, data collected from user input at
control stations, and survey responses. Ultimately, it is
intended that users will be able to upload their own recordings, which may then be edited and indexed according to
criteria derived from the experiments.
10. REFERENCES
[1] F. Friedman, A. Naparstek, and M. Taussig-Rubbio,
”Alarmingly Useless: The Case for Banning Car
Alarms in New York City,” Transportation Alternatives, 2003.
[2] N. Misdariis, A. Gruson, and P. Susini, ”Detectability study of warning signals in urban background
noises: a first step for designing the sound of electric
vehicles,” in Proceedings of Meetings on Acoustics
(POMA)-ICA 2013, Montreal, 2013, pp. 1-9.
[3] G. Lemaitre, P. Susini, S. Winsberg, B. Letinturier,
and S., McAdams, ”The sound quality of car horns:
Designing new representative sound,” Acta acustica
united with Acustica, vol. 95, no. 2, pp. 356-372,
2009.
[4] C. Suied, P. Susini, and S. McAdams, ”Evaluating Warning Sound Urgency With Reaction Times,”
Journal of Experimental Psychology: Applied, vol.
14, no. 3, pp. 201-212, 2008.
[5] P. Susini, N. Misdariis, G. Lemaitre, and O. Houix,
”Naturalness influences the perceived usability and
pleasantness of a user interface’s sonic feedback,”
Journal on Multimodal User Interfaces, vol. 5, no.
3, pp. 52-80, 2012.
[6] P. Schaeffer, Traité des objets musicaux. Paris,
France: Seuil, 2002.
[10] G. Peeters and X. Rodet, ”Automatically selecting
signal descriptors for Sound Classification,” International Computer Music Conference Proceedings,
Gteborg, Sweden, 2002, pp. 455-458.
[11] R.D. Patterson, ”Auditory warning sounds in the
work environment,” in Philosophical Transactions of
the Royal Society of London B: Biological Sciences,
1990, pp. 485-492.
[12] J. Edworthy, E. Hellier, K. Titschener, A. Naweed,
and R. Roels, ”Heterogeneity in auditory alarm sets
make them easier to learn,” International Journal of
Industrial Economics, vol. 41, no. 2, pp. 135-146,
2011.
[13] A. Minard, N. Misdariis, O. Houix, and P. Susini,
”Catégorisation des sons environnementaux sur la
base de profiles morphologiques,” 10ème Congrès
Français d’Acoustique, Lyon, France, 2010, pp.1-6.
[14] K. Yamauchi, J. Choi, R. Maiguma, M. Takada, and
S. Iwamiya, ”A Basic Study on Universal Design of
Auditory Signals in Automobiles,” Journal of Physiological Anthopology and Applied Human Science,
vol. 23, no. 6, pp. 295-298, 2004.
[15] S. Kuwano, S. Namba, A. Schick, H. Höge, H. Fastl,
T. Filippou, and M.Florentine, ”Subjective impression of auditory danger signals in different countries,” Acoustical Science and Technology, vol. 28,
no. 5, pp. 360-362, 2007.
[16] R.D. Patterson and A.J. Datta, ”Extending the Domain of Auditory Warning Sounds: Creative Use of
High Frequencies and Temporal Asymmetry,” in Human Factors in Auditory Warnings, N. Stanton and
J. Edworthy (Eds.), Brookfield, VT, USA: Ashgate,
1999, pp. 73-88.
[17] N. Stanton and J. Edworthy, ”Auditory affordances
in the intensive treatment unit,” Applied Ergonomics,
vol 29, no. 5, pp. 389-394, 1998.
[18] N. Misdariis, A. Minard, P. Susini, G.Lemaitre, S.
McAdams, and E. Parizet, ”Environmental Sound
Perception: Metadescription and Modeling Based on
Independent Primary Studies,” EURASIP Journal on
Audio, Speech, and Music Processing, January 2010,
Article No. 6.
[19] G. Peeters, A large set of audio features for sound description (similarity and classification), CUIDADO
project Ircam technical report, 2004.
– 23–
Journal of the Japanese Society for Sonic Arts, Vol.1 No.1 pp.17–21
11. AUTHORS’ PROFILES
Alexander Sigman
Alexander Sigman’s award-winning instrumental, electroacoustic, multimedia, and installation works have been featured on major international festivals, exhibitions, institutions, and venues across Europe, Asia, Australia, and the
US. In June 2007, Sigman was Composer-in-Residence
at the Musiques Dmesures festival in Clermont-Ferrand,
France. Subsequently, he was awarded residency fellowships by the Akademie Schloss Solitude (Stuttgart, Germany), the Djerassi Foundation, and the Paul Dresher Ensemble Artists Residency Center. In 2013-2014, he undertook a musical research residency at IRCAM.
Sigman completed his doctorate in Music Composition at Stanford University in 2010. Prior to Stanford,
he obtained a BM in Music and a BA in Cognitive Sciences from Rice University. Further postgraduate studies
were undertaken at the University for Music and the Performing Arts Vienna, as well as the Institute for Sonology of the Royal Conservatory in The Hague (Netherlands). He is currently Assistant Professor of Composition
at Keimyung University in Daegu, South Korea. More information may be found here: www.lxsigman.com.
Nicolas Misdariis
icolas Misdariis is a research fellow and the co-head of Ircam/Sound Perception and Design team. He is graduated
from a college of university level, in 1993, specializing
in professional training on mechanics (CESTI-SupMeca).
He made a specialization in acoustics within the Acoustical Laboratory of Maine University (LAUM, Le Mans).
Since 1995, he has worked at Ircam as researcher where
he has taken part in several projects concerning different
fields of research dealing with sound science and technology. In 1999, he contributed to the constitution of the Ircam/Sound Design team where he has mainly developed
works related to sound synthesis, diffusion technologies,
environmental sound and soundscape perception, auditory
display, and interactive sonification. Since 2010, he is also
a lecturer within the Master of Sound Design at the Fine
Arts school at Le Mans.
– 24–